System architecture and method of processing data therein

1. A heuristic method of reducing energy consumption in a system having a plurality of subsystems, comprising:

identifying one or more approximation parameters in each of a plurality of subsystems of a system;

for an application that is run on the system with a predefined quality minimum:

a) reducing performance of each of the plurality of subsystems by identifying a corresponding value for each of the one or more approximation parameters in each of the plurality of subsystems such that quality of the application is about the same as the predefined quality minimum;

b) determining energy savings for the system based on setting the one or more approximation parameters to the corresponding values for each of the plurality of subsystems while maintaining the other of the plurality of subsystems at non-reduced performance;

c) sorting the plurality of subsystems based on system-level energy savings;

d) classifying each of the plurality of subsystems into coarse and fine subsystems, whereby changing values for the one or more approximation parameters of a coarse subsystem results in substantially more energy savings than changing values for the one or more approximation parameters of a fine subsystem; and

e) optimizing the values of the one or more approximation parameters while maintaining quality level of the application at or above the predefined quality minimum by i) adjusting the values of the one or more parameters of the coarse subsystems, and ii) adjusting the values of the one or more parameters of the fine subsystems.

2. The method of claim 1 , wherein the optimization method is by a gradient descent algorithm.

3. The method of claim 1 , wherein the coarse and fine subsystems classification is based on energy saving impact of about 2 to 1 for coarse subsystems.

4. The method of claim 1 , wherein the coarse and fine subsystems classification is based on energy saving impact of about 3 to 1 for coarse subsystems.

5. The method of claim 1 , wherein the coarse and fine subsystems classification is based on energy saving impact of about 5 to 1 for coarse subsystems.

6. The method of claim 1 , wherein the coarse and fine subsystems classification is based on energy saving impact of about 10 to 1 for coarse subsystems.

7. The method of claim 1 , wherein the coarse and fine subsystems classification is based on energy saving impact of about 50 to 1 for coarse subsystems.

8. The method of claim 1 , wherein the steps a)-e) are performed at an initial calibration stage of running the application.

9. The method of claim 8 , wherein the steps a)-e) are repeated periodically after the initial calibration stage.

10. The method of claim 1 , wherein the energy savings of the system is between about 7.5 times to about 14.4 times of the system with no approximation for a quality degradation of about 1% to about 7.5%, respectively.

11. A computing system comprising of a plurality of subsystems, comprising:

a memory subsystem including a non-transitory computer readable medium;

a computing subsystem including a processor, the processor configured to implement a heuristic method of reducing energy consumption in the system having the plurality of subsystems, comprising:

identifying one or more approximation parameters in each of the plurality of subsystems of the system;

for an application that is run on the system with a predefined quality minimum:

a) reducing performance of each of the plurality of subsystems by identifying a corresponding value for each of the one or more approximation parameters in each of the plurality of subsystems such that quality of the application is about the same as the predefined quality minimum;

b) determining energy savings for the system based on setting the one or more approximation parameters to the corresponding values for each of the plurality of subsystems while maintaining the other of the plurality of subsystems at non-reduced performance;

c) sorting the plurality of subsystems based on system-level energy savings;

d) classifying each of the plurality of subsystems into coarse and fine subsystems, whereby changing values for the one or more approximation parameters of a coarse subsystem results in substantially more energy savings than changing values for the one or more approximation parameters of a fine subsystem; and

e) optimizing the values of the one or more approximation parameters while maintaining quality level of the application at or above the predefined quality minimum by i) adjusting the values of the one or more parameters of the coarse subsystems, and ii) adjusting the values of the one or more parameters of the fine subsystems.

12. The computing system of claim 11 , wherein the optimization method is by a gradient descent algorithm.

13. The computing system of claim 11 , wherein the coarse and fine subsystems classification is based on energy saving impact of about 2 to 1 for coarse subsystems.

14. The computing system of claim 11 , wherein the coarse and fine subsystems classification is based on energy saving impact of about 3 to 1 for coarse subsystems.

15. The computing system of claim 11 , wherein the coarse and fine subsystems classification is based on energy saving impact of about 5 to 1 for coarse subsystems.

16. The computing system of claim 11 , wherein the coarse and fine subsystems classification is based on energy saving impact of about 10 to 1 for coarse subsystems.

17. The computing system of claim 11 , wherein the coarse and fine subsystems classification is based on energy saving impact of about 50 to 1 for coarse subsystems.

18. The computing system of claim 11 , wherein the steps a)-e) are performed at an initial calibration stage of running the application.

19. The computing system of claim 18 , wherein the steps a)-e) are repeated periodically after the initial calibration stage.

20. The computing system of claim 11 , wherein the energy savings of the system is between about 7.5 times to about 14.4 times of the system with no approximation for a quality degradation of about 1% to about 7.5%, respectively.